pyhf

pyhf copied to clipboard

Summary

For quick checks, discovery significance calculation via the CLI would be helpful to have.

Additional Information

There are (at least) two ways of going about this. The pyhf cls interface could almost be used as-is, but currently does not support q0 as test statistic:

https://github.com/scikit-hep/pyhf/blob/34dd45c84426c5e14a280b7da8742d0d4fd7c83d/src/pyhf/cli/infer.py#L143

With q0 added there, the following works:

pyhf cls --test-stat q0 --test-poi 0.0 some_workspace.json 

It might be confusing to use something called cls for the discovery significance calculation. Instead it may be better to leave the cls CLI for CLs calculations and have a separate section for discovery significance.

Code of Conduct

• [X] I agree to follow the Code of Conduct

For quick checks, discovery significance calculation via the CLI would be helpful to have.

Seems very reasonable. :+1:

It might be confusing to use something called cls for the discovery significance calculation. Instead it may be better to leave the cls CLI for CLs calculations and have a separate section for discovery significance.

This would be my preferred way of doing it. We should try to design APIs that try to distinguish the statistical/inference concepts being used.